The dark side of NoSQL

Re:The dark side of NoSQL by cturner on Wednesday September 30, 2009 @07:38AM (#852224)

I've had an unusual problem with couchdb. A bad query can flatten your host. The engine uses every bit of performance in pursuit of its answer, which might be the workload equivalent of a series of stupid open joins in SQL. Your bash consoles become unresponsive. The same mistake with a traditional relational database causes less damage because those databases are less effective at harvesting the power of the host.

There's a training obstacle. I know how SQL works and how to make SQL perform well almost by instinct. NoSQL requires me to relearn data querying. The lack of standardisation across offerings means that the knowledge I acquire will be less transferable what I've accumulated from Oracle/Postgresql/Openbase/Mysql/Sqlite.

[ Parent | on HN ]

Re:The dark side of NoSQL by grumpycanuck on Wednesday September 30, 2009 @09:19AM (#852351)

I don't think that the people who use CouchDB (or similar non-traditional data stores) are really worried about "tranferable knowledge". They're just interested in solving a problem they think the NoSQL group of technologies can help them with.

[ Parent | on HN ]

Re:The dark side of NoSQL by Maro on Wednesday September 30, 2009 @10:02AM (#852418)

You should think of CouchDB as the Web2.0 equivalent of Microsoft Access. It's good for running a simple web-app off, because you can store the HTML, JS and the data all in CouchDB (I think). So one use-case may be a simple billing system for your local surf shop that is sure as hell not going to store a lot of data or handle a lot of load. I got this impression from a Damien Katz talk. (I think the use-case they use is a recipe app.)

[ Parent | on HN ]

Re:The dark side of NoSQL by AndrewO on Wednesday September 30, 2009 @01:49PM (#853144)

> you can store the HTML, JS and the data all in CouchDB

I think you're talking of CouchApp, which allows you to create apps like that on top of CouchDB. Of course, if that was the maximum of what one was trying to do, one could get similar functionality from an RDBMS, some simple routing rules, and a CGI script that served content straight from the DB.

I guess my point is that there's not really anything intrinsic to CouchDB that makes this possible: it's a pretty short and inviting conceptual leap from "this DB does everything through an HTTP interface" to "I can run my entire application from this DB". My worry (and I'm not accusing the parent of this) is that that misses out on all of the other things that CouchDB offers (especially the ability to merge DBs that have been disconnected for long periods, which often gets lost behind the HTTP interface and MapReduce hand-waving).

[ Parent | on HN ]

Re:The dark side of NoSQL by logicalmind on Wednesday September 30, 2009 @05:58PM (#853819)

The difference being that if you want to answer a question like "what hour of the day is my store busiest?" you would have to write a MapReduce function in Erlang for couchdb. Where alternatively, anyone with minimal knowledge can figure out that query in SQL.

[ Parent | on HN ]

Re:The dark side of NoSQL by tumult on Wednesday September 30, 2009 @07:03PM (#853934)

You mean, they would have to write a simple JavaScript function. No Erlang involved in writing CouchDB views.

[ Parent | on HN ]

Re:The dark side of NoSQL by peregrine on Wednesday September 30, 2009 @12:38PM (#852902)

Well with the load balancer, and continuous replication you could theoretically scale to any level easily.

[ Parent | on HN ]

Re:The dark side of NoSQL by Maro on Wednesday September 30, 2009 @01:21PM (#853075)

You've got something that is inherently an order of magnitude slower than other solutions. Why would you bother to try to scale it? It's not the right tool for the job.

[ Parent | on HN ]

Re:The dark side of NoSQL by jbellis on Wednesday September 30, 2009 @12:58PM (#852995)

No. CouchDB doesn't support partitioning, so all your replicas need to be able to handle the full write volume.

[ Parent | on HN ]

Re:The dark side of NoSQL by boorad on Thursday October 01, 2009 @10:52AM (#855004)

It will soon. We have an 8-node cluster running in PROD right now. A Dynamo-like layer provides consistent hashing, partitioning, read/write quorum constants and the like.

[ Parent | on HN ]